PHYSICAL REVIEW B 80, 144506 共2009兲

Josephson scanning tunneling microscopy: A local and direct probe of the superconducting order parameter Hikari Kimura,1,2,* R. P. Barber, Jr.,3 S. Ono,4 Yoichi Ando,5 and R. C. Dynes1,2,6,† 1 Department of Physics, University of California, Berkeley, California 94720, USA Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA 3Department of Physics, Santa Clara University, Santa Clara, California 95053, USA 4Central Research Institute of Electric Power Industry, Komae, Tokyo 201-8511, Japan 5 Institute of Scientific and Industrial Research, Osaka University, Ibaraki, Osaka 567-0047, Japan 6Department of Physics, University of California–San Diego, La Jolla, California 92093-0319, USA 共Received 10 August 2009; revised manuscript received 15 September 2009; published 6 October 2009兲

2

Direct measurements of the superconducting superfluid on the surface of vacuum-cleaved Bi2Sr2CaCu2O8+␦ 共BSCCO兲 samples are reported. These measurements are accomplished via Josephson tunneling into the sample using a scanning tunneling microscope 共STM兲 equipped with a superconducting tip. The spatial resolution of the STM of lateral distances less than the superconducting coherence length allows it to reveal local inhomogeneities in the pair wave function of the BSCCO. Instrument performance is demonstrated first with Josephson measurements of Pb films followed by the layered superconductor NbSe2. The relevant measurement parameter, the Josephson ICRN product, is discussed within the context of both BCS superconductors and the high transition temperature superconductors. The local relationship between the ICRN product and the quasiparticle density of states 共DOS兲 gap are presented within the context of phase diagrams for BSCCO. Excessive current densities can be produced with these measurements and have been found to alter the local DOS in the BSCCO. Systematic studies of this effect were performed to determine the practical measurement limits for these experiments. Alternative methods for preparation of the BSCCO surface are also discussed. DOI: 10.1103/PhysRevB.80.144506

PACS number共s兲: 74.40.⫹k, 74.25.Dw, 74.50.⫹r, 74.72.Hs

I. INTRODUCTION

Scanning tunneling microscopy 共STM兲 and scanning tunneling spectroscopy 共STS兲 have been extensively utilized for nanometer scale studies of physical and electronic structures on the surface of high transition temperature 共high-TC兲 superconducting cuprates, especially Bi2Sr2CaCu2O8+␦ 共BSCCO兲 and YBa2Cu3O7−␦ 共YBCO兲. A rich array of experiments includes imaging vortex flux lines1–3 and mapping the integrated local density of states 共LDOS兲.4–6 These latter results were used to determine a gap 2⌬ taken to be the difference between two coherence peaks in the LDOS. The spatial resolution combined with spectroscopic results has been used to produce “gap maps” over the surface for various dopings in BSCCO. These studies have revealed that 共i兲 the formation of gapped regions obtained from the dI / dV spectra actually start above TC, and there is a linear relation between ⌬ and the gap opening temperature, Tⴱ 共Ref. 7兲; and 共ii兲 there is an anticorrelation between the energy gap ⌬ in the superconducting state and the normal state conductance at the Fermi energy8 for various dopings of BSCCO. A challenge with this approach is that the identification of ⌬ becomes ambiguous in strongly underdoped BSCCO where the observed dI / dV curves no longer have well-defined sharp coherence peaks.9,10 Although these STM/STS experiments on high-TC superconductors have yielded a wealth of data, they suffer from important limitations. Of particular significance in our opinion is that because they utilize a normal metal tip, these studies can only probe the local quasiparticle density of states 共DOS兲. This DOS almost certainly has an intimate connection with the superconductivity in these materials; 1098-0121/2009/80共14兲/144506共16兲

however, that relationship is still unknown. Furthermore, the derived gap qualitatively changes its shape with doping. In contrast, the BCS theory for conventional superconductors defines a well-established relationship between the gap in the quasiparticle DOS 共⌬BCS兲 and fundamental quantities of the superconducting state including the amplitude of the order parameter and the superconducting transition temperature, TC. Without a similar theory for the high-TC superconductors to enable the inference of the superconducting properties from the quasiparticle DOS, it is necessary to directly probe the superconducting superfluid of these materials. Two central questions that such a direct probe should address are 共i兲 whether the superconducting order parameter of BSCCO has spatial variation and 共ii兲 how the superconducting ground state correlates with the quasiparticle excited states 共⌬兲. In this paper we will present the results of experiments using an STM with a superconducting tip. This instrument allows us to directly probe the superconducting superfluid at the surface of BSCCO using the Josephson effect. We will first discuss the primary quantity derived from Josephson tunneling measurements, the ICRN product, and its relationship to the fundamental properties of the superconductors that make up the tunnel junction. After a description of the technical aspects of the apparatus, data which verify its successful operation on a conventional superconductor 共Pb兲 will be presented. The approach is then extended to a layered superconductor 共NbSe2兲 and then finally to the high-TC superconductor BSCCO. BSCCO is believed to be primarily a d-wave superconductor but with a weak s-wave component. Furthermore, the STM probes the order parameter at the surface where the symmetry restrictions might be relaxed somewhat. This experiment relies on coupling to the s-wave component.

144506-1

©2009 The American Physical Society

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al. II. ICRN PRODUCT

Josephson tunneling is Cooper pair tunneling between two superconductors separated by a thin barrier. The zerovoltage supercurrent flowing through the junction is given by I = IC sin共␸2 − ␸1兲, where IC is the maximum zero temperature supercurrent that the junction can sustain and ␸1共2兲 is the phase of two superconducting electrodes’ order parameter. The maximum supercurrent is related to the amplitude of the superconducting pair wave function. An STM with a superconducting tip can be a local Josephson probe and can, in principle, access the superconducting pair wave function directly on a length scale smaller than or comparable to the superconducting coherence length, ␰. Between the two superconductors in a tunnel junction, the Josephson ICRN product 共RN is the normal state resistance of the junction兲 is a directly measurable quantity uniquely determined by the specific materials. ICRN is a fundamental parameter that is directly linked to the superconducting order parameter amplitude 兩⌿兩, and in the case of conventional superconductors to the energy gaps ⌬BCS through the BCS relationship.11 Josephson studies using a superconducting STM on conventional superconductors have shown good agreement between the measured ICRN and BCS predictions.12,13 For high-TC superconductors, on the other hand, there is no established theory to relate ICRN with ⌬ derived from the quasiparticle excitation spectrum. There can be at least two additional issues not present for conventional superconductors. The first is that we do not have a universally agreed-upon theory of the high-TC cuprates and second, the symmetry of the order parameter has a substantial dx2−y2 component thus making the coupling to a conventional superconductor 共s-wave symmetry兲 more complicated. An ICRN measurement on BSCCO using a conventional superconducting STM should, however, both prove the existence and yield the amplitude of the BSCCO pair wave function that couples to the conventional superconducting tip. Because of the spatial resolution of an STM, this measurement could reveal useful information regarding inhomogeneities in the superconductivity of BSCCO. III. EXPERIMENT A. Superconducting tip

A reproducible and stable superconducting tip fabrication method that we have developed begins with a Pt0.8 / Ir0.2 tip mechanically cut from a 0.25 mm diameter wire.14 Tips are then placed in a bell-jar evaporator with the tip apex pointing toward the evaporation sources. 5500 Å of Pb is deposited at a rate of ⬃40 Å / s followed by 36 Å of Ag at a rate of 1 Å / s without breaking vacuum. The thick layer of Pb was chosen such that at 2.1 K, well below transition temperature of Pb 共TC = 7.2 K兲, there would be bulk superconductivity in the tip 共superconducting coherence length, ␰0, of Pb is ␰0 = 830 Å兲. The Ag serves as a capping layer to protect the Pb layer from rapid oxidation upon exposure to the atmosphere. The Ag layer is thin enough to proximity couple to the Pb layer resulting in a superconducting tip with TC and ⌬BCS only slightly below that of bulk Pb. Because the Pb/Ag bi-

layer is deposited without breaking vacuum, the interface between the layers is expected to be clean, and thus superconductivity may be induced in the Ag layer by the proximity effect.15 The Pb/Ag combination is also a good metallurgical choice, as there is no significant alloying at the interface.16 For Josephson measurements of a conventional superconductor, Pb/Ag samples were also evaporated onto freshly cleaved graphite substrates during the same deposition as the tips. The same Ag capping layer keeps these samples stable for the transfer from the evaporator to the STM apparatus. B. NbSe2

Single-crystal 2H-NbSe2 was chosen as a first target material beyond conventional superconductors as a surrogate to the high-TC superconducting cuprates. 2H-NbSe2, a family of layered transition-metal dichalcogenides, is a type-II conventional superconductor with TC = 7.2 K and charge-density wave 共CDW兲 transition at TCDW = 33 K, so the CDW state coexists with the superconducting state below TC. This material also has short coherence lengths 共␰a,b = 77 Å and ␰c = 23 Å兲, an anisotropic s-wave gap varying from 0.7 to 1.4 meV across the Fermi surface17 and multiband superconductivity indicated from observations of momentum-dependent superconducting gap on the different Fermi surface sheets.18 Van der Waals bonding between Se layers is so weak that the crystal is easily cleaved to expose a fresh and inert surface for STM measurements. However, no Josephson tunneling measurements have been reported before those of our group13 partly because it is difficult to grow a stable insulating 共usually oxide兲 layer for planar tunnel junctions. C. Bi2Sr2CaCu2O8+␦

There is still much discussion and controversy about the symmetry of the order parameter of high-TC superconducting cuprates and whether the pseudogap state observed in the underdoped region in BSCCO is a precursor to the coherent superconducting state. If the symmetry is strictly d-wave, there should exist no Josephson coupling between high-TC superconducting cuprates and a conventional superconducting tip. However, Josephson coupling has been observed between conventional superconductors and YBCO 共Ref. 19兲 as well as BSCCO.20,21 Thus, Josephson measurements of BSCCO using a superconducting STM should reveal results about the symmetry of the BSCCO order parameter. Correlations between ICRN products from the Josephson effect and the energy gap ⌬ measured from the quasiparticle excitation spectra of BSCCO should contribute to the construction of a microscopic theory of the mechanism of high-TC superconducting cuprates. An apparent constraint indicated by a previous study of BSCCO at high STM currents5 that we will discuss later is that there is a threshold current above which the BSCCO morphology and spectroscopy are drastically and irreversibly changed. This effect appears at a tunneling current of about 500 pA. Care will be taken to sweep the bias of the STM junction such that the resulting tunneling current does not exceed this threshold to avoid changing the BSCCO electronic structure.

144506-2

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

PHYSICAL REVIEW B 80, 144506 共2009兲

BSCCO single crystals are grown by the floating zone method22 with the hole doping, ␦h, ranging between heavily underdoped 共TC = 64 K兲 and overdoped 共TC = 74 K兲 via optimally doped samples 共TC = 94 K兲. TC was determined by magnetic susceptibility measurements. In these current studies, extensive Josephson measurements were performed on overdoped samples with different dopings 共TC = 76, 79, and 81 K兲. Our BSCCO samples which have typical dimensions of 1 mm⫻ 1 mm and a few tens of ␮m thick are glued onto a copper plate with silver epoxy 共Epoxy technology EE 129– 4兲. All of the samples were cleaved in high vacuum 共⬃3 ⫻ 10−8 Torr兲 at room temperature to expose a fresh surface. BiO-BiO planes are attracted by weak van der Waals bonding so that they are most likely the cleavage plane,23 with the conducting CuO2 surface two layers below. Cleaved samples are then cooled to T = 2.1 K after being inserted into the STM. The experiments are performed at 2.1 K where the properties of Pb approach the low temperature limit. While lower temperatures would nominally reduce the fluctuations, we find that a characteristic noise temperature Tn of the apparatus dominates the fluctuations. A lower base temperature in the current configuration would therefore not improve the measurements; only a significant reduction in the noise leakage from room-temperature electronics would do so.

0

D. EJ, phase fluctuations, Tn, current density

The signature Josephson response of a superconducting STM differs from that of typical low RN planar S/I/S devices. For identical superconductors, Ambegaokar and Baratoff11 derived the temperature-dependent Josephson binding energy, EJ, where EJ共T兲 =

冉 冊

⌬共T兲 បIC ␲ប ⌬共T兲 tanh . = 2 2e 4e RN 2kBT

共1兲

Because of the experimental base temperature 共T = 2.1 K兲 and large RN associated with an STM, EJ is smaller than kBT for the Josephson STM. For example, with an STM resistance of 50 k⍀, E j / kB is roughly 1 K. Also for ultrasmall tunnel junctions, the Coulomb charging energy EC can be large. We estimate the capacitance, C, of the STM junction formed between the conical tip apex and the sample surface to be about 1 fF. EC = e2 / 2C is therefore of order 1 K: comparable to both EJ and kBT. The time scale of an electron tunneling in the STM junction and conducting off the tip is much shorter than ប / EC, so that the electron is swept away long before the charging effects become relevant. Because kBT is the dominant energy, the phase difference of the two superconductors, ␸, is not locked in a minimum of the sinusoidal EJ vs ␸ washboard potential but is thermally excited and diffusive 共Fig. 1兲. With a current bias, the phase diffuses preferentially in one direction as illustrated in Fig. 1. Near zero-bias voltage the observed Josephson current is therefore dependent on the bias due to the dissipative phase motion. The phase diffusion model was first proposed by Ivanchenko and Zil’berman24 and further developed by others.25,26 In this model we can consider the thermal fluctuations as Johnson noise generated by a resistor ZENV at a noise temperature Tn; both parameters depend only on the

2π 4π 6π 8π 10π

ϕ

kBTn

U(ϕ) EJ

ωP FIG. 1. Josephson phase dynamics of the washboard potential in the classical thermal fluctuation regime.

experimental setup. In the limit of ␣ = EJ / kBTn ⱕ 1, they derived a simple analytic form for the I-V characteristics of the thermally fluctuated Josephson currents, I共V兲 =

IC2 ZENV V . 2 V2 + V2P

共2兲

As described above, the relevant energy scale of our STM Josephson junctions is ␣ ⱕ 1, so that the analytic form of Eq. 共2兲 is applicable to analyzing our data. Now the pair current has a voltage dependence due to the diffusive phase motion. V P as a function of Tn and ZENV is the voltage where the pair current becomes maximum. V P will not change if Tn and ZENV are constant parameters intrinsic to the junction’s environment, while IC increases as RN is decreased. Thus, the thermally fluctuated Josephson current is characterized by three quantities, maximum supercurrent IC, Tn, and ZENV. Tn is an effective noise temperature for the ultrasmall junction. This temperature can be elevated by noise from the room-temperature electronics unless all the leads connecting to the junction are heavily filtered. ZENV is the impedance of the junction’s environment or the electronic circuit where the junction is embedded. It is reported that for ultrasmall tunnel junctions, the Josephson phase dynamics is at very high frequency characterized by the Josephson plasma frequency, ␻ P or EJ / ប 共for STM Josephson junctions, it is of order of 1011 ⬃ 1012 Hz兲 and the frequency-dependent damping at this frequency region is dominated by stray capacitance and inductance of the cables connecting the junction to the external circuit.27–30 The cables will load the junction with an impedance on the order of the free space impedance, Z0 = 冑␮0 / ␧0 = 377 ⍀. Experimentally we are interested in determining the Josephson ICRN product, a quantity characteristic of the superconductivity of the constituent materials. If we observe the phase diffusion branches in our STM Josephson junctions, we can characterize them by identifying the two parameters, Tn and ZENV, and then we can directly derive ICRN of the material of interest by comparing the observed data and fits to the phase diffusion model. Another concern for implementing a Josephson STM is the high current density due to the small geometry of the STM junction. A low junction resistance is desired such that EJ is maximized, but it results in a high current density. In this configuration the current density, j, can be calculated using the tunnel current I = 10 nA which is a typical value for Pb/I/Pb STM Josephson junctions and the effective diameter of the superconducting tip, ⬃3 Å 共for fcc structure of

144506-3

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

FIG. 2. Normalized dI / dV spectrum of a Pb/I/Pb STM junction at T = 2.1 K. The Pb phonon structures can be seen as dips at 7 and 12 mV: energies corresponding respectively to the transverse and longitudinal phonon energies in Pb as measured from the energy gap edge.

Ag, the nearest-neighbor distance is 2.89 Å兲 over which electrons are being injected. This calculation yields j ⬃ 107 A / cm2, a very high current density. Nevertheless, as presented later, Josephson current was observed using the SC STM tip and no self-heating effect due to the high current density was observed for Pb and NbSe2.31 For BSCCO, however, previous work reported that the high current density caused a huge effect on both its electronic structure and morphology.5

FIG. 3. I-V characteristics of Pb/I/Pb STM junctions at T = 2.1 K. I-V curves are measured as the junction normal state resistance RN is varied 共z position兲 with the tip position 共x , y兲 fixed. At these high resistances, EJ is too small for the Josephson effects to be observable. Also Coulomb blockade effects are negligible because the resistance of the tip and the film is such that the charge is swept away on a short time scale compared with EC.

Figure 4 shows the observed I-V characteristics at lower voltages 共lines兲 for STM Josephson junctions formed between a Pb/Ag superconducting tip and a Pb/Ag superconducting film. The top panel of Fig. 4 shows that the location

IV. EXPERIMENTAL RESULTS A. Pb

To test the operation of our SC-STM, we first studied Ag-capped Pb films with our Ag-capped Pb tip 共a symmetric junction兲. The spectrum we obtained is shown in Fig. 2 and is characteristic of S/I/S tunnel junctions: very sharp coherence peaks corresponding to eV = ⫾ 共⌬tip + ⌬sample兲, from which we obtained ⌬tip = ⌬sample = 1.35 meV, slightly smaller than the bulk value for Pb 共⌬bulk = 1.4 meV兲 due to the proximity effect of the Ag capping layer. Moreover, the deviations from the BCS density of states outside the Pb gap due to strong-coupling effects are clearly seen at energies corresponding to the transverse and longitudinal phonon energies, eVT − 2⌬ = 4.5 meV and eVL − 2⌬ = 8.5 meV, respectively. Furthermore, just above the large coherence peaks we see the effects of the Ag proximity on the superconducting Pb. This shows up as a small dip just above the peak at 2⌬. Figure 3 presents several I-V curves measured at different RN by changing the tip-sample distance sequentially. The superconducting gap size remains unchanged as RN is decreased. Low leakage current below the Pb gap 共shown in Fig. 3兲 and the observation of the phonon structure in Fig. 2 confirm high quality vacuum tunnel junctions. To acquire these I-V characteristics, the tunnel current feedback loop is temporarily turned off only during the measurement. We find that in the time required to do these measurements and the Josephson measurements later, the tip position remains stable. The feedback is re-established immediately after each I-V acquisition.

FIG. 4. I-V characteristics of Pb/I/Pb STM junctions at T = 2.1 K. 共Top panel兲 Apparent are both the onset of the tunnel current at V = 2⌬ and structures near zero bias 共inside the box兲. 共Bottom兲 I-V characteristics near zero bias for lower junction resistances than those in the top frame. The lines display the measured thermally fluctuated Josephson current and the symbols represent two-parameter fits to the phase diffusion model.

144506-4

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 5. 共a兲 Plot of IC ⫻ 冑e / kBTn vs GN of Pb/I/Pb STM junctions. The slope is equal to ICRN ⫻ 冑e / kBTn and is shown as a linear fit. From the fitted slope and using the known value of ICRN共Pb/ I / Pb兲, Tn and ZENV are determined to be 15.9⫾ 0.1 K and 279⫾ 9 ⍀, respectively. 共b兲 Log-log plot of IC ⫻ 冑e / kBTn vs GN for data from three different preamplifiers 关including that from 共a兲兴. All data fall onto the same single line indicating Tn does not change due to large current flowing through the tunnel junction. These results give us confidence in using these two parameters Tn and ZENV later in our determination of ICRN for NbSe2 and BSCCO.

of the gap does not change as RN is lowered. The bottom panel is a close-up view of I-V characteristics near zero bias and clearly shows peaked structures first appearing and then exhibiting increasing heights as RN is decreased 共EJ enhanced兲. These observed I-V curves are fitted to the phase diffusion model 关Eq. 共2兲兴 with two parameters, V P and IC. The best fits to the phase diffusion model are represented by the symbols in the bottom panel of Fig. 4 and the quality of these fits convince us that we have observed the signature of pair tunneling. This analysis yields a plot of IC ⫻ 冑e / kBTn vs GN = 1 / RN, expected to be linear with zero intercept 共no IC at infinite RN兲 and a slope equal to ICRN ⫻ 冑e / kBTn, as shown in Fig. 5共a兲. We can calculate ICRN of Pb from the Ambegaokar-Baratoff formula11 or Eq. 共1兲 using ⌬ = 1.35 meV at T = 2.1 K and including a factor of 0.788 due to strong electron-phonon coupling in Pb.32 For T = 2.1 K and TC 共Pb兲 = 7.2 K, the hyperbolic tangent is very close to unity and we get ICRN 共Pb/ I / Pb兲 = 1.671 mV. Substituting this value into the slope of the linear data fit in Fig. 5共a兲 for our STM Josephson junctions, we can determine Tn and ZENV, which are parameters depending only on the experimental setup. Current values of these quantities for our

apparatus are 15.9⫾ 0.1 K and 279⫾ 9 ⍀, respectively. The fact that Tn is higher than the 2.1 K base temperature can be explained by leakage of rf noise to the junction from roomtemperature electronics. This value for Tn is an improvement from our earlier work as a result of low temperature filters inserted close to the microscope. Further improvements are called for as Tn is still higher than 2.1 K. ZENV is close to the expected value of the impedance of free space as described above. These quantities were measured using a preamplifier with 109 gain. We repeated the Josephson measurements to derive Tn and ZENV using other preamplifiers with lower gains 共107 and 108兲 to cover the larger tunneling current range. All three data sets of the ICRN plots in Fig. 5共b兲 fall on the same single line, convincing us that the Josephson coupling is enhanced as RN is decreased and Tn and ZENV are constant parameters intrinsic to the experimental circumstances and configuration. This also indicates that Tn remains the same even for high current density 共lower RN兲. No heating effects or degraded superconducting DOS of the Pb tip or sample were observed. The lowest junction normal state resistance we studied was 4.6 k⍀, which is smaller than the quantum resistance of a single channel in the ballistic regime, RQ = h / 2e2 = 12.9 k⍀. Subharmonic gap structures observed in the Pb/ I/Pb STM Josephson junctions indicate that the low resistance STM junction is not in the weak tunneling limit 共兩T兩2 = D = 10−6 or smaller兲, but somewhat higher transparency, D ⬃ 0.1, contributed from several conduction channels.33 B. NbSe2

The data of Pb/ I / NbSe2 STM Josephson junctions are presented in Fig. 6. The experimental data on the bottom panel 共lines兲 are contributions from the thermally fluctuated Josephson currents after subtraction of the quasiparticle background due to thermally excited quasiparticles. The background is obtained from the I-V curves of high resistance junctions where no Josephson currents are observed. Figure 6 shows good agreement between the fits to the phase diffusion model 共symbols兲 and the observed data. Since the slope value in Fig. 7 is equal to ICRN ⫻ 冑e / kBTn and we can assume that Tn and ZENV remain constant, we can write a 冑kBTn / e = ICRN共Pb兲 / slope共Pb兲 relationship = ICRN共Pb/ NbSe2兲 / slope共Pb/ NbSe2兲, where slope共Pb兲 is obtained from the linear fit to the data in the IC ⫻ 冑e / kBTn vs GN plot of the Pb/I/Pb STM Josephson junctions shown in Fig. 5. Substituting the known values, ICRN共Pb兲 / slope共Pb兲 and the measured slope共Pb/ NbSe2兲, we obtain ICRN共Pb/ NbSe2兲 = 1.39⫾ 0.03 mV. We then use the formula for the Josephson binding energy for different superconductors at T = 0 K given the gaps of each34 EJ =

冉

冊

兩⌬1 − ⌬2兩 ប ⌬ 1⌬ 2 K , e RN ⌬1 + ⌬2 ⌬1 + ⌬2 2

共3兲

where K共x兲 is a complete elliptic integral of the first kind. Substituting ⌬1 = ⌬Pb = 1.35 meV and ⌬2 for the smallest and average gap of NbSe2, that is, 0.7 and 1.1 meV, respectively, into Eq. 共3兲 yields 1.34 mV⬍ ICRN共Pb/ NbSe2 , T = 0 K兲

144506-5

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

FIG. 8. Optimally doped BSCCO topography scanned by superconducting STM tip at T = 2.1 K. The white bar is equal to 10 Å.

FIG. 6. I-V characteristics of Pb/ I / NbSe2 STM junctions at T = 2.1 K. 共Top panel兲 Apparent is a current rise at V = ⌬Pb + ⌬NbSe2. 共Bottom兲 I-V characteristics near zero bias for lower junction resistances than those in the top frame showing thermally fluctuated Josephson current. The symbols represent two-parameter fits to the phase diffusion model.

⬍ 1.70 mV. Our result of 1.39 mV is in good agreement with the theoretical expectation.

C. Bi2Sr2CaCu2O8+␦

Figure 8 is an atomic resolution image of cleaved optimally doped BSCCO scanned by the superconducting STM tip at T = 2.1 K. Because of the thick Pb layer used in our superconducting tip fabrication, it is difficult to routinely obtain atomic resolution images. However, we can easily locate

FIG. 7. Plot of IC ⫻ 冑e / kBTn vs GN for Pb/ I / NbSe2 STM junctions. The slope is equal to ICRN ⫻ 冑e / kBTn and is shown as a linear fit to the data. From the fitted slope and using the known value of Tn = 15.9 K, ICRN共Pb/ NbSe2兲 is determined.

step edges and isolate flat surfaces where all the present data were obtained. S/I/S STM junctions formed between the superconducting Pb tip and overdoped BSCCO single crystals show different features than those observed for the Pb/I/Pb STM junctions. First, the energy gap of BSCCO is an order of magnitude larger than the Pb gap and second, the dI / dV spectrum for the BSCCO gap has a “gaplessness”—nonzero conductance at the Fermi energy—with an asymmetric normal state background conductance. Figure 9共a兲 presents an I-V characteristic of Pb/I/overdoped BSCCO STM junctions at T = 2.1 K taken at RN = 10 M⍀ clearly showing the Pb gap around 1.4 meV. The Pb gap edge does not have a sharp onset of tunnel current compared to that of Pb/I/Pb STM junctions because states exist all the way to the Fermi energy in the density of states of BSCCO. In other words, quasiparticles can tunnel at the Fermi energy of BSCCO. The inset of Fig. 9共a兲 shows a dI / dV spectrum in the region of the Pb gap. The conductance outside the Pb gap is affected by the large energy gap of BSCCO 共⌬BSCCO = 40 meV as measured by the energy of the coherence peak.兲. Figure 9共b兲 shows a dI / dV spectrum taken with a large sweep range for the local density of states of BSCCO at the same location 共RN = 500 M⍀兲. For the local Josephson measurements for BSCCO single crystals, we first observe the dI / dV spectrum at a particular surface point on overdoped BSCCO 共TC = 79 K兲 in order to measure the energy gap ⌬ 共solid line in the inset of Fig. 10兲. We use standard lock-in techniques with 1 kHz modulation and a 2.5 mVRMS modulation voltage on the bias voltage and a junction normal resistance, RN ⬃ 500 M⍀. Although it is clearly a simplification of a more complex structure, we use the same definition for ⌬ as in previous works35 in order to make comparisons. We then decrease RN to enhance EJ in order to observe the pair tunnel current. A difference is that we cannot use very large currents with BSCCO due to the current limits for BSCCO damage 共to be discussed below兲. This limitation also impacts the determination of RN. Since the energy gap of BSCCO is much larger than that of Pb it is more difficult to measure RN from the I-V curves because of the limits on maximum current. Our procedure is to record several I-V curves by sweeping the bias voltage above the Pb

144506-6

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 10. I-V characteristics of Pb/I/overdoped BSCCO 共TC = 79 K兲 STM Josephson junctions at T = 2.1 K. The Pb gap is clearly seen around V = 1.4 mV. Inset: dI / dV spectrum 共solid line兲 measured before low RN measurements showing sharp coherence peaks with ⌬ = 37 meV. dI / dV spectrum measured after low RN measurements 共dotted line兲 indicates an LDOS change due to high current density.

FIG. 9. 共a兲 I-V characteristic of Pb/I/overdoped BSCCO STM junctions at T = 2.1 K clearly showing a Pb gap around 1.4 meV. Note the absence of leakage although the Pb gap edge is smeared compared to that of Pb/I/Pb STM junctions due to finite states all the way to the Fermi energy in the BSCCO density of states. Inset: dI / dV for the Pb gap taken at RN = 10 M⍀. The conductance outside the Pb gap is affected by a relatively large energy gap of BSCCO 共⌬BSCCO = 40 meV兲. 共b兲 dI / dV spectrum taken over a large voltage range for the BSCCO gap at the same location. RN = 500 M⍀. The modulation amplitude added to the bias voltage is 2.5 mVRMS.

gap. Then RN is determined by making use of our knowledge that the junction normal resistance inside the BSCCO gap 共above the Pb gap兲 is 3–4 times larger than that outside the BSCCO gap determined from the dI / dV spectrum such as illustrated in the inset of Fig. 10. For the lower junction resistance I-V curves, RN was calculated from the factor required to scale the current so that it overlaps with already normalized I共V兲RN versus V curves with the ratio of the conductance inside to that outside the BSCCO gap. A set of dI / dV data for the BSCCO gap and I-V curves taken with progressively lower junction resistances are necessary to calculate the ICRN product every time the tip is moved to a new location. In the main frame of Fig. 10 we plot the I-V characteristics at lower bias and lower RN. A low leakage current below the Pb gap confirms the high quality of the vacuum tunnel junctions. Further decreasing RN increases the quasiparticle tunneling probability and finally the contribution from the thermally fluctuated Josephson currents is observed when EJ is comparable to kBTn.36 Figure 11 displays a close-up view of the I-V characteristics near zero bias clearly showing that the superconducting Pb tip was Josephson coupled to the BSCCO. The quasipar-

ticle background represented by the dotted line in Fig. 11共b兲 is obtained from an average of several normalized I-V curves at higher RN. At high RN there is no contribution from the Josephson effect and we use this curve as background. We scale it to the RN of the lower resistance data and the difference shown in Fig. 11共b兲 is due to the Josephson currents.

FIG. 11. 共a兲 Low bias I-V characteristics of Fig. 10 for various junction resistances at T = 2.1 K. 共b兲 Averaged I-V characteristic near zero bias for quasiparticle background 共dotted line兲. One of the observed I-V curves is shown by the solid line. 共c兲 Thermally fluctuated Josephson currents peaked at V P as derived by subtracting quasiparticle background 关Fig. 11共b兲兴 from the I-V curves 关Fig. 11共a兲兴. The data are represented by the lines and the symbols represent two-parameter fits to the phase diffusion model.

144506-7

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

FIG. 12. Plot of IC ⫻ 冑e / kBTn vs GN of Pb/I/overdoped BSCCO 共TC = 79 K兲 STM Josephson junctions. The slope is equal to ICRN ⫻ 冑e / kBTn. Using the fitted slope and substituting the previously determined Tn, the Josephson product at this surface point is found to be ICRN = 335 ␮V.

Figure 11共c兲 shows the remaining contributions from the thermally fluctuated Josephson current after subtracting the quasiparticle background of Fig. 11共b兲 from the I-V curves of Fig. 11共a兲. The data in Fig. 11共c兲 are shown as lines and the best fits to the Eq. 共2兲 are represented by the symbols. These good fits convince us that we have likely observed the pair current between a conventional 共s-wave兲 superconducting Pb tip and overdoped BSCCO. This suggests that the BSCCO does not have a pure d-wave order parameter at least at the surface. In addition, the dI / dV spectrum represented by the dotted line in the inset of Fig. 10 was observed after the lowest RN measurements in Fig. 11. The LDOS has changed significantly during the measurements and the quasiparticle coherence peaks have disappeared perhaps due to the high current density of the measurements at the highest conductance studied. This “modified” dI / dV curve resembles those previously observed in heavily underdoped BSCCO,7,9,10 in the “pseudogap” state at temperatures above TC 共Refs. 7 and 37兲, and in strongly disordered BSCCO thin films.38 It is also similar to the dI / dV spectra observed by others on surfaces which were altered by scanning with large tunnel currents.5 It is important to note that LDOS changes were observed only after measurements were made with RN below 30 k⍀ and I above the threshold current around 500 pA. Moreover, the Josephson current disappeared after these irreversible changes in the LDOS on BSCCO occurred. In order to avoid this effect, most of the data presented here were obtained with RN ranging from 30 to 100 k⍀. This effect will be discussed in subsequent sections. Each fit to the Josephson portion of the I-V curves in Fig. 11共c兲 generates a single data point in the plot shown in Fig. 12 in the similar way as described in the Pb/I/Pb and Pb/ I / NbSe2 STM junction results. As GN is increased 共RN is reduced兲 the observed IC increases 共EJ increases兲. We now rely on the values for Tn and ZENV that we determined from our measurements on Pb/I/Pb for this experimental apparatus, and taking the slope of the linear fit shown in Fig. 12, we find ICRN at this surface point to be 335 ␮V. There are numerous normal tip STM studies of BSCCO aimed at developing a spatial picture of the electronic quan-

FIG. 13. I-V characteristics of Pb/I/overdoped BSCCO 共TC = 79 K兲 STM Josephson junctions at different location from that in Fig. 10. 共a兲 I-V characteristics of Pb/I/overdoped BSCCO 共TC = 79 K兲 STM Josephson junctions at T = 2.1 K. Inset: dI / dV measured before low RN measurement 共black line兲 and after it 共gray line兲. Note that energy gap is unchanged. 共b兲 Thermally fluctuated Josephson currents 共lines兲 and fits 共circles兲 to the phase diffusion model. 共c兲 Plot of IC ⫻ 冑e / kBTn vs GN; ICRN = 279 ␮V. Two data points at GN = 15 and 39 ␮S correspond to I-V characteristics without any pair current observed.

tities of this material. Several of these investigations have produced renderings referred to as gap maps.35 These images reveal the inhomogeneous nature of the energy gap6 and periodic electronic modulation both inside the vortex cores39 and above TC.40 Again these data were derived from quasiparticle excitation spectra, not probing the superconducting pair state itself. It is natural to ask 共i兲 whether the superconducting order parameter of BSCCO has spatial variation, and 共ii兲 how the superconducting ground state correlates with the quasiparticle excited states 共⌬兲. Since we have the capability to measure both ⌬ 共Fig. 10兲 and ICRN 共Fig. 11兲 at the same location on the surface, we have used these techniques to address these questions. In order to avoid the irreversible change in the LDOS for higher currents 共Fig. 10 inset兲, the minimum junction resistance was kept above 30 k⍀. The result for a second location on overdoped BSCCO 共TC = 79 K兲 is presented in Fig. 13. Unlike the case shown in Fig. 10, the inset of Fig. 13共a兲 shows no appreciable change in the LDOS after the I-V measurement at the lowest RN. We found ⌬ = 47 meV at this surface point. ICRN derived from

144506-8

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 14. I-V characteristics and ICRN plot of Pb/I/overdoped BSCCO 共TC = 76 K兲 STM Josephson junction for studying reproducibility of IC. 共a兲 I-V curve 共solid line兲 measured before low RN measurements showing ⌬ = 40 meV. I-V curve measured after low RN measurements 共dotted line兲 indicating that the energy gap at this surface point remained the same. 共b兲 Plot of IC ⫻ 冑e / kBTn vs GN. The number labels represent the chronological order of the data sets. This order clearly shows the Josephson currents disappearing 共5兲 and reappearing 共6 and 7兲. Data points of 共1,2兲 and 共6,7兲 were measured repeatedly at RN = 80 and 63 k⍀, respectively.

these data is also different from the previous location. Of note in Fig. 13共c兲 are two data points showing zero IC appearing in the ICRN plot. The disappearances of IC are observed at GN = 15 and 39 ␮S 共I-V curves without any Josephson contributions兲. In order to check reproducibility of the pair current, we first repeated the Josephson current measurement at the same RN as that when currents had been observed. We also checked the linear relationship between IC and GN 共both increasing and decreasing GN兲. Reproducibility was usually observed. However, occasionally IC disappeared. Figure 14 shows the ICRN plot of Pb/I/overdoped BSCCO 共TC = 76 K兲 measured at T = 2.1 K. We first measured the I-V characteristic for the BSCCO gap 关solid line in Fig. 14共a兲兴. Then RN was decreased to see the Josephson current on the I-V characteristics. In Fig. 14共b兲, each Josephson measurement is labeled in the chronological order in which it was taken. The Josephson coupling increases as we decrease RN 共expected兲 but then unexpectedly disappears at a lower RN 共label 5兲. Increasing RN 共decreasing GN兲 results in the Josephson coupling returning 共labels 6 and 7兲. This behavior is all unexpected. After these low RN measurements, the large bias I-V curve was again recorded to observe the BSCCO gap. This curve 共dashes兲 in Fig. 14共a兲 indicates that

FIG. 15. Spatial studies of both 共a兲 ⌬ and 共b兲 ICRN at the same locations on overdoped BSCCO 共TC = 79 K兲. ICRN changes spatially and seems to anticorrelate with ⌬. Note that no Josephson contributions were observed at the surface points denoted by arrows 关in 共b兲兴 where the largest energy gaps were measured in this 20 Å ⫻ 25 Å region. 共c兲 The line cut of dI / dV spectra is measured along the y axis in 共a兲 showing a well-known gap inhomogeneity 共offset for clarity兲.

the LDOS of the BSCCO and the energy gap remains the same before and after the low RN measurements so that the disappearance of IC was not caused by the LDOS change due to high current density. Although the origin of this disappearance is still under investigation, we observe that low RN measurements 共RN below ⬃300 k⍀兲 on BSCCO increase the low frequency noise on the tunnel current. This noise appears to be induced locally on the BSCCO and not from the environment or the electronics. These effects were not ever seen in our Josephson studies on the Pb-Pb or NbSe2 systems. We assure ourselves that we have not affected the tip during the measurements by verifying that the Pb gap is always reproduced and the exponential decrease in the tunnel current vs the tip-sample distance is also observed after low RN measurements. Keeping these observations in mind, we performed both local Josephson and spectroscopic measurements on the overdoped BSCCO surface 共TC = 79 K兲. Figure 15 shows the spatial dependence of the energy gap and ICRN measured simultaneously every 5 – 10 Å on a particular region on the surface. It clearly indicates that ICRN and therefore the superconducting pair wave function of BSCCO changes on a nanometer-length scale on the surface. More interestingly, we can see an anticorrelation between ⌬ and ICRN such that

144506-9

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

ICRN tends to be reduced as ⌬ increases. This is not predicted by the BCS theory where ⌬BCS and ICRN are linearly correlated. The tendency was also observed in the data taken along a line of 100 Å on another overdoped BSCCO 共TC = 76 K兲 sample. This relationship is more apparent when ICRN is plotted in the next section as a function of ⌬ for a variety of experiments. V. DISCUSSION A. d-wave superconductors

The observation of c-axis Josephson coupling in planar Pb-YBa2Cu3O7−␦ 共YBCO兲 single-crystal Josephson junctions has been reported and was explained by an s-wave component in the order parameter of YBCO induced by an orthorhombic distortion.19 Although the crystallographic symmetry of Bi2Sr2CaCu2O8+␦ 共BSCCO兲 makes s- and d-wave mixing less likely,41 Josephson coupling in planar junctions between conventional superconductors and BSCCO has been observed.20,21 ICRN values for these junctions 共Nb- or PbBSCCO兲 ranged from 1 to 10 ␮V suggesting that the s component is about three orders of magnitude smaller than the d component. Because ICRN was measured in macroscopic junctions in previous work, any strong local inhomogeneities were obscured and meaningful comparisons with an inhomogeneous ⌬ could not be made. It is, therefore, very important to locally probe the order parameter in this strongly inhomogeneous material using Josephson tunneling. For high-TC superconducting cuprates where the pairing mechanism is still under debate, attempts to extract the possible coupling due to the strong electron-phonon interaction were done for YBa2Cu3O7−␦ although the authors cautioned that the gap observed in the normalized conductance data for YBa2Cu3O7−␦ is not of the BCS form. Nevertheless, the expected TC was calculated from the normal state parameters, ␣2F共␻兲 and ␭ derived from the observed dI / dV spectrum and found that it was 2/3 of the measured TC of this material.42 Recently microscopic studies of the phonon structure by STM were performed for Bi2Sr2CaCu2O8+␦ 共Ref. 43兲 and electron-doped cuprate, Pr0.88LaCe0.12CuO4.44 They have, however, extracted the phonon energies from positive peaks of d2I / dV2 spectra, with an assumption that the observed gap was equal to the superconducting gap rather than following the previous procedure. Furthermore, it has been suggested that their results could be interpreted as inelastic tunneling associating with apical oxygen within the barrier.45,46 Electrons tunneling from the STM tip can lose energy to an oxygen vibrational phonon mode inside the barrier yielding a new tunneling channel and mimicking the results reported. Furthermore, angle-resolved photoemission spectroscopy 共ARPES兲 data have been interpreted within the context of a strong electron-phonon interaction model.47 B. Variation in ICRN in overdoped Bi2Sr2CaCu2O8+␦

We interpret these results within the framework of the phase diagram for high-TC superconducting cuprates proposed by Emery and Kivelson.48 High-TC superconducting

FIG. 16. Phase diagram based on the phase fluctuation model of high-TC superconductors as functions of temperature T and hole doping, ␦h, proposed by Emery and Kivelson. Optimally doped 共OP兲 region is a crossover from underdoped 共UD兲 to overdoped 共OV兲 region. The phase ordering temperature, T␪, and the meanfield transition temperature, Tⴱ, are defined in the text.

cuprates are doped Mott insulators with low superfluid density, nS. Therefore, the phase stiffness, which is the energy scale to twist the phase, is small in these superconductors such that phase fluctuations could play an important role in determining TC. There are two possible temperature scales that could affect the transition to superconductivity. T␪ is a temperature at which the phase ordering disappears because the phase stiffness disappears. Another temperature scale, Tⴱ, is described as the temperature below which a gap on the quasiparticle spectrum appears. On the low doping side, a system could be divided into regions where the order parameter is well defined locally but not globally 共a granular superconductor where the grains are weakly coupled to each other兲. These areas becomes larger with increasing doping resulting in stronger intergranular coupling so that the phase coherence length becomes longer and less susceptible to phase fluctuations. T␪ is increased as the hole doping ␦h increases leading to a rise of the global TC of the sample. Meanwhile, Tⴱ continues to decrease as ␦h increases. At the optimal doping, T␪ and Tⴱ cross over, so that the whole sample region is now phase coherent but the mean-field value 共energy gap兲 of the sample is suppressed. Thus, T␪ and Tⴱ are the upper bounds to TC. T␪ is more significant due to the phase fluctuations on the lower doping side 共underdoped兲, while Tⴱ is more important on the higher doping side 共overdoped兲. Figure 16 plots TC vs hole doping, ␦h, based on this Emery-Kivelson model and the superconducting region forms a dome shape with the maximum TC at ␦h ⬃ 0.16 共optimally doped兲. Decreasing or increasing ␦h from this value results in a TC decrease. In order to interpret our results using this Emery-Kivelson model 共TC vs ␦h兲, we make two assumptions in order to replace ␦h in their model by the energy gap, ⌬, which we actually measure in our experiments.36 First of all, Tⴱ changes monotonically with ␦h and decreases as ␦h increases. Previous STM studies9,10 reported the spatially averaged gap value monotonically increased from overdoped 共the average ⌬ ⱕ 40 meV兲 to underdoped 共the average ⌬ ⱖ 60 meV兲, and dI / dV with ⌬ ⱖ 65 meV is often observed in heavily underdoped samples to no longer exhibit sharp coherence peaks. We also measured three samples with different dopings: un-

144506-10

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 17. Typical dI / dV spectra and the corresponding averaged energy gaps, ⌬AVE, at T = 2.1 K for BSCCO with three different dopings. They are underdoped 共TC = 64 K兲, optimally doped 共TC = 94 K兲, and overdoped 共TC = 76 K兲 samples. Note that ⌬AVE monotonically decreases as ␦h increases.

derdoped 共TC = 64 K兲, optimally doped 共TC = 94 K兲, and overdoped 共TC = 76 K兲 and observed this tendency as shown schematically in Fig. 17. The results indicate that the average ⌬, ⌬AVE, seems to monotonically increase as ␦h is decreased. It was reported that the formation of gapped regions obtained from the dI / dV spectra actually started above TC and there is a linear relation between ⌬ and the gap opening temperature, Tⴱ, for optimally doped and overdoped BSCCO samples.7 Combining with all these facts, we suggest that the ␦h axis in the Emery-Kivelson model can be transformed into the ⌬ axis, but now Tⴱ monotonically increases with ⌬AVE as shown in Fig. 18. Second, McElroy et al.9 reported that all the gap-map studies for different dopings, ranging from underdoped to overdoped, show not only strong gap inhomogeneity over all samples, overdoped or underdoped 共observation of the larger gap in regions of the overdoped and that of the smaller gap in regions of the underdoped samples兲, but also the shape of the averaged dI / dV spectra for a given gap seems to be very similar, independent of whether the bulk sample is overdoped or underdoped. These lead us to make the second assumption that although the bulk 共macroscopic兲 doping of each BSCCO sample is characterized by the transport TC and the spatially averaged energy gap, ⌬AVE, local doping which will determine the local superconducting nature of the sample 共the locally measured ⌬, the pair amplitude, TC兲 reflects the observed inhomogeneity. Putting it another way, the smaller gap region which is sparsely distributed on the underdoped sample behaves as “overdoped,” while the larger gap region which is rarely observed in overdoped sample behaves as “underdoped.” This suggestion is also supported by the recent finding of local Fermi surface variations on BSCCO indicating that local doping is not equivalent to the macroscopic doping of the sample.49 Since we measure ⌬, we choose to replot the EmeryKivelson model schematically as shown in Fig. 18. We flip the TC vs ␦h relation in the Emery-Kivelson model to a TC vs ⌬AVE relation. Thus, the region, where the smaller gap is

FIG. 18. Modified Emery-Kivelson model which includes two assumptions: 共1兲 the linear relation between ⌬ and Tⴱ and 共2兲 a local doping variation on the BSCCO surface.

measured, we regard as an overdoped region, while the region with larger gap is regarded as underdoped region, even though our local Josephson measurements are done on overdoped samples 共TC = 76, 79, and 81 K兲. Now Tⴱ 共and delta兲 monotonically increases and the dome-shaped region is simply flipped horizontally as shown when plotted vs ⌬. With this adaptation in hand, we now summarize our measurements of the Josephson ICRN product vs ⌬ for five overdoped samples with each data point taken at locations roughly 5 – 10 Å apart.36 Shown in Fig. 19 is the summary plot with the modified Emery-Kivelson model superimposed. Although there is scatter in the observed ICRN values for a given ⌬, this figure clearly indicates the nanometer scale inhomogeneities in both ICRN and ⌬. The reason for the scatter from experiment to experiment is under investigation. We believe this scatter is related to the microscopic inhomogeneities of BSCCO. We do not see this kind of scatter in the investigations of Pb/Ag or NbSe2. A consistent but surprising feature seen from this plot is that ICRN tends to be a maximum when ⌬ is between 40 and 45 meV, and the trend is for it to decrease or become zero as ⌬ increases or decreases from this maximal point. Our results in Fig. 19 show that ICRN is maximized at a gap value of 40–45 meV, the average ⌬ typically observed in optimally doped BSCCO 共corresponding to the highest TC samples兲. ICRN decreases as ⌬ becomes larger. It also decreases as ⌬ becomes smaller. The ICRN vs ⌬ that we measure behaves in a similar way as TC vs ␦h 共as ␦h changes

144506-11

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

FIG. 19. ICRN vs ⌬ overlaid with the Emery-Kivelson model. Sketches of TC and Tⴱ from the Emery-Kivelson model are shown by dotted and dashed lines, respectively. The vertical scale for the model curves is arbitrary 共obtained from Ref. 36兲.

toward zero from the critical doping ⬃0.3 at the end of the superconducting region兲, and it does not follow the behavior of ⌬ vs TC as expected from BCS theory. It is important to reiterate that for any given sample, we observed inhomogeneities both in ⌬ and ICRN values as a function of location. From our results we correlate the observed ICRN with the amplitude of the superconducting order parameter 兩⌿兩 as well as with the TC of BSCCO via the Emery-Kivelson model phase diagram.36 On the underdoped side of the phase diagram, these three quantities 共ICRN, 兩⌿兩, and TC兲 decrease 共smaller superfluid density兲 as ⌬ increases and anticorrelate with Tⴱ. This inverse relation between ICRN and ⌬ in BSCCO is an unconventional result because in the BCS picture ⌬BCS, ICRN, 兩⌿兩, and TC are all correlated. On the overdoped side of the phase diagram, TC decreases as ␦h is increased above 0.16 共⌬ also becomes smaller in this doping region; a conventional result兲. Since the overdoped side is the amplitude dominated region, Tⴱ closely relates to ⌬ and hence decreases as ␦h is increased. ICRN decreases as ⌬ is decreased from the value around 40 meV in Fig. 19 indicating ICRN, 兩⌿兩, TC, ⌬, and Tⴱ behave similarly and conventionally as ␦h is increased toward the critical doping 共␦h ⬃ 0.3兲 where TC vanishes. Another possible framework for discussing our results is the two-gap scenario observed in recent ARPES measurements.50,51 In the underdoped regime in BSCCO samples, a smaller energy gap is observed and it becomes larger with increasing doping in the nodal region distinct from the larger energy gap 共pseudogap兲 in the antinodal region where no coherence peaks are observed. Moreover, a temperature dependence of the nodal gap follows the BCS functional form very well, while the antinodal gap remains finite at TC. The same trend of ⌬共T兲 is also observed in overdoped samples, but the gapless region above TC expands on the Fermi surface with increasing doping 共suppression of the pseudogap兲. It has also been reported that two gaps are observed in overdoped 共Bi1−yPby兲2Sr2CuO6+x using a variable temperature STM.52 It was claimed that there are smaller homogeneous energy gaps vanishing near TC as measured from the dI / dV spectra normalized by normal state conductance, as well as the larger inhomogeneous energy

gaps, which have very weak temperature dependence. Although consistent with our ICRN measurements, we do not observe the second gap directly. To our knowledge, the momentum k component of the tunneling electron parallel to the junction barrier is conserved in the tunnel process, but the very small confinement of the electron due to the STM tip might increase the uncertainty of momentum and relax the constraint for momentum conservation. Thus, the tunneling current observed in the STM could possibly be averaged over a large fraction of the momentum space, therefore, making it difficult to resolve a momentum-dependent gap by STM. Moreover, the results in Fig. 19 represent measurements of both ICRN and ⌬ averaged over momentum space, and therefore we are unable to address this alternate model. C. Current density effect

We have observed notable changes in the LDOS after low RN 共high current density兲 measurements. Most of these observations involve preliminary results and more studies are necessary to come to a quantitative conclusion. Nevertheless, we have observed this effect so often that it deserves reporting. The effect is illustrated in the inset of Fig. 10. Several questions arise: Does the LDOS change suddenly or continuously? When or at what RN does it happen? How does LDOS change relate to the Josephson current? We performed “back and forth” measurements in which the BSCCO gap 共high RN measurement兲 was measured followed by low RN measurements at the same location on the surface. Results are shown in Fig. 20. We first measured dI / dV labeled 1, then measured I-V characteristics at lower voltages in the region of the Pb gap. RN is lowered until it reaches that of the I-V curve labeled 7. The tip is backed up to increase RN to measure dI / dV for the BSCCO gap labeled 8 and so on. It is interesting that the BSCCO gap remains almost unchanged after measuring the Pb gap at RN = 68 k⍀, but a large LDOS change is observed 共dI / dV labeled 20兲 after obtaining the I-V curve labeled 19 at RN = 11 k⍀. The dI / dV curve labeled 20 indicates not only a disappearance of sharp coherence peaks but an apparent increase in the energy gap size. Further decreasing RN to measure the Pb gap 共I-V curve labeled 24兲 makes the LDOS change to a “V” shape 共dI / dV labeled 25兲 where we can no longer define an energy gap. From similar measurements, we have observed that the BSCCO gap and the shape of dI / dV rarely change by measuring the Pb gap until RN is reduced to around 30 k⍀, but further decreasing RN causes a deformation of LDOS. Howald et al. observed qualitatively similar dI / dV curves on the intentionally disordered surface by scanning with large tunnel current.5 Their tunnel condition, however, was at I = 500 pA with V = −200 mV so the power dissipated from the tip was 10−10 W, while the typical tunnel condition used in these current measurements for the I-V curve at RN = 30 k⍀, for example, is I = 500 pA with V = 1.2 mV so that the power dissipated from the superconducting tip is 100 times smaller than that used by Howald et al. although tip-sample distance in our STM junctions is smaller. In this configuration the current density, j, could be calculated using the tunnel current I = 500 pA and the effective diameter of the superconducting

144506-12

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 21. dI / dV spectra taken along a line from the originally damaged surface point 共offset for clarity兲. The bottom curve is the same dI / dV spectrum labeled 25 in Fig. 20共b兲. D. Different surface preparation

FIG. 20. 共a兲 I-V characteristics and 共b兲 dI / dV spectra taken at the same surface point on overdoped BSCCO 共TC = 74 K兲 at T = 2.1 K. LDOS change in this sample is caused by low RN measurements. The numbers labeled for I-V and dI / dV curves are measured in chronological order. For example, dI / dV spectrum 共14兲 is taken right after I-V curve 共13兲 was measured.

tip, ⬃3 Å, over which electrons are being injected. Thus, it yields j ⬃ 106 A / cm2, a very high current density. It is still under investigation to answer why RN ⬃ 30 k⍀ is the threshold resistance for the BSCCO’s LDOS change. We conclude that current density is the relevant parameter that causes these surface and spectral changes. We have also measured the lateral range of the high current density alteration of the LDOS by moving the tip away from the original altered location to measure dI / dV curves a few nanometers away. Figure 21 shows the degraded LDOS is continuously changing, finally recovering to the superconducting LDOS as the tip is moved away from the originally damaged point. The dI / dV spectrum with sharp coherence peaks is recovered at 13 nm away from the damaged point in opposite directions along a line. While Howald et al. “burned” the surface by scanning with a large tunnel current, we burned at a specific surface point by taking I-V characteristics at low RN. It is interesting to note that spatial destruction of the BSCCO superconducting LDOS is similar for both experiments. Only qualitative studies of the high current density effect on BSCCO’s electronic structure have been done so far, however, a relation between the thermally fluctuated Josephson current and LDOS change is still unknown. Further study is necessary to discuss this effect quantitatively.

Cleaving the BSCCO to expose an atomically flat surfaced is widely used for STM studies of this material; however, no study of the effect of the cleaving on the electronic structure of BSCCO has been reported. This fact results in a question as to whether the gap inhomogeneity routinely observed on cleaved surfaces is intrinsic to BSCCO or a result of the cleave. It is well known that the superconducting tunneling probes the depth of a coherence length into the sample surface. Therefore it is important to address this question because the electronic degradation on the surface of BSCCO could affect its tunneling current due to the very short c-axis coherence length 共␰C ⱕ 1 nm兲 compared with much longer ␰C of conventional superconductors. Chemical etching is an alternate method to remove a degraded surface layer and possibly make a passivated layer. A chemical etching technique, originally reported by Vasquez et al.53 was applied to Pb/I/YBCO tunnel junctions54,55 and Josephson junctions.19 An STM study of etched YBCO single crystals56,57 revealed that etching with 1% bromine 共Br兲 by volume in methanol resulted in an etching rate of 250 Å / min. The etching proceeds layer by layer and results in large flat areas separated by steps with single unit cell depth 共⬃12 Å兲. The etching also produced pits on the surface which expand radially, introducing some surface roughness, but further etching removed layers without increasing roughness. For BSCCO single crystals, we observed that 1% Br in methanol was strong enough to dramatically roughen the etched surface causing the STM to tip crash. This result suggests that BSCCO is more sensitive to the etching than YBCO. In order to optimize the etching condition, etching rate calibrations were performed as follows: the BSCCO single crystal was coated with thinned rubber cement leaving a small region of exposed BSCCO. The etching solution consisting of 0.1% Br in methanol was kept on the BSCCO single crystal for 3 min. The surface was then rinsed by dipping the sample in methanol in an ultrasound cleaner and the rubber cement was removed by sonication in toluene. A commercial profiler revealed clear steps at various edges of the etched region of roughly 6000 Å in height. This result

144506-13

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al.

FIG. 22. STM image of 0.1% Br etched overdoped BSCCO 共TC = 74 K兲 at room temperature. Scan size is 800 Å ⫻ 800 Å.

FIG. 24. Room-temperature STM image of the same surface in Fig. 22 with larger scan area 共3200 Å ⫻ 3200 Å兲.

indicates etch rates for BSCCO using 0.1% Br of 2000 Å / min. Figure 22 shows a typical surface of overdoped BSCCO 共TC = 74 K兲 etched in 0.1% Br in methanol for 3 min followed by sonication in methanol and finally blown dry with nitrogen gas. The etched surface consists of “pancakes” with lateral dimensions approximately a few hundred angstroms. These pancakes have various step heights of not only a half unit cell in depth, 15 Å 共Fig. 23兲, but also 5 and 10 Å. Figure 24 shows a large area scan of the same sample as Fig. 22. A vertical corrugation over this surface is less than 30 Å indicating that the 0.1% Br etching proceeds layer by layer. Figure 25 shows the dI / dV spectra measured on the 0.1% Br etched overdoped BSCCO sample at T = 4.2 K. Two dI / dV curves were taken 10 Å apart. It is noteworthy that the spectral line shape looks very similar to that observed on the cleaved BSCCO sample except the asymmetry typically observed for cleaved samples in the coherence peaks is reversed. Nevertheless, the result is reproducible. This sample was then cooled to T = 2.1 K to measure the Pb gap. Figure 26 shows the I-V characteristic measured at a lower RN. The Pb gap was clearly seen around V = 1.4 mV although the I-V curve was not taken at the same surface point as Fig. 26. Further investigation is required to determine whether the gap inhomogeneity is intrinsic to this material. It would also be useful to study how the dI / dV spectra and the Josephson ICRN product vary over an etched surface.

It is shown from these preliminary experiments that the Br etching proceeds layer by layer on BSCCO and yields a passivated surface. Observations of the superconducting Pb gap ensure that the passivated layer is thin enough for vacuum tunneling; however, difficulty in reproducible observation of the Pb gap suggests that the thickness might change from etched sample to sample. Low temperature image scans also appear noisier than room-temperature images and there is difficulty at obtaining reproducible low temperature images. Since Br must be handled under a fume hood because of its high volatility and toxic nature, the etching process is done in air. This constraint possibly results in surface contamination although it still remains puzzling why the image is better 共less noisy and reproducibly obtained兲 at room temperature than at low temperature. In summary, we have prepared the BSCCO surface by chemical etching for our superconducting STM study. The etched BSCCO surface yields reproducible dI / dV spectra; however, extensive low RN measurements to observe Josephson current have not been accomplished yet.

FIG. 23. Surface height cross section along the line shown in Fig. 22.

VI. CONCLUSION

We have described a series of experiments that utilize a superconductor-tipped STM to perform Josephson tunneling

FIG. 25. dI / dV spectra with a large bias measured on the 0.1% bromine etched overdoped BSCCO at T = 4.2 K. Two dI / dV spectra were measured 10 Å apart.

144506-14

PHYSICAL REVIEW B 80, 144506 共2009兲

JOSEPHSON SCANNING TUNNELING MICROSCOPY: A…

FIG. 26. I-V characteristic measured at lower RN on the 0.1% bromine etched overdoped BSCCO at T = 2.1 K. The Pb gap is clearly seen around V = 1.4 mV.

measurements on the high-TC superconducting cuprate BSCCO. These measurements are motivated by the desire to directly access the pair wave function given the lack of a theory to connect the quasiparticle DOS to the superconducting state, in contrast to BCS superconductors. Operation of the apparatus has been verified by measurements of a conventional superconductor, Pb, and then the layered superconductor NbSe2. We find good agreement between these measurements and theoretical predictions giving us confidence in our BSCCO data for which no prediction is available. Our results indicate that like the quasiparticle DOS, the pair wave

function is also inhomogeneous over the doping range studied, with variations on length scales of roughly 1 nm. Furthermore, we find that the gap measured from the quasiparticle DOS is anticorrelated with the Josephson ICRN product for areas where the local superconducting nature has the characteristics and is consistent with underdoped samples. In addition we observe that excessive current densities can irreversibly alter the LDOS in these samples and we have determined a limit on the current. Taking care to stay below those limits allows us to avoid this effect. In an effort to determine whether the local inhomogeneities are intrinsic or the result of the surface preparation by cleaving, we have also performed measurements of BSCCO samples that have been etched with a Br/methanol solution. Although no Josephson signal has been detected, this approach does appear to yield reasonable surfaces, albeit with apparent tip contamination challenges. ACKNOWLEDGMENTS

We thank the Berkeley Physics Machine shop for expert technical assistance and Garg Assoc Pvt Ltd. for providing low noise miniature coax cable. The work in Berkeley was supported by DOE Grant No. DE-FG02-05ER46194. S.O. was supported by KAKENHI Grant No. 20740213 and Y.A. by KAKENHI Grants No. 19674002 and No. 20030004.

and A. Baratoff, Phys. Rev. Lett. 10, 486 共1963兲. Naaman, W. Teizer, and R. C. Dynes, Phys. Rev. Lett. 87, 097004 共2001兲. 13 O. Naaman, R. C. Dynes, and E. Bucher, Int. J. Mod. Phys. B 17, 3569 共2003兲. 14 O. Naaman, W. Teizer, and R. C. Dynes, Rev. Sci. Instrum. 72, 1688 共2001兲. 15 A. S. Katz, A. G. Sun, R. C. Dynes, and K. Char, Appl. Phys. Lett. 66, 105 共1995兲. 16 M. Hansen, Constitution of Binary Allows, 2nd ed. 共McGrawHill, New York, 1958兲. 17 H. F. Hess, R. B. Robinson, and J. V. Waszczak, Physica B 169, 422 共1991兲. 18 T. Yokoya, T. Kiss, A. Chainani, S. Shin, M. Nohara, and H. Takagi, Science 294, 2518 共2001兲. 19 A. G. Sun, D. A. Gajewski, M. B. Maple, and R. C. Dynes, Phys. Rev. Lett. 72, 2267 共1994兲. 20 M. Mö␤le and R. Kleiner, Phys. Rev. B 59, 4486 共1999兲. 21 I. Kawayama, M. Kanai, M. Maruyama, A. Fujimaki, and H. Hayakawa, Physica C 325, 49 共1999兲. 22 S. Ono and Y. Ando, Phys. Rev. B 67, 104512 共2003兲. 23 S. H. Pan, E. W. Hudson, and J. C. Davis, Appl. Phys. Lett. 73, 2992 共1998兲. 24 Y. M. Ivanchenko and L. A. Zil’berman, Zh. Eksp. Teor. Fiz. 55, 2395 共1968兲 关Sov. Phys. JETP 28, 1272 共1969兲兴. 25 G.-L. Ingold, H. Grabert, and U. Eberhardt, Phys. Rev. B 50, 395 共1994兲. 26 Y. Harada, H. Takayanagi, and A. A. Odintsov, Phys. Rev. B 54, 6608 共1996兲.

*Present address: Department of Physics and Astronomy, Univer-

11 V. Ambegaokar

sity of California, Irvine, California 92697, USA. † [email protected] 1 I. Maggio-Aprile, C. Renner, A. Erb, E. Walker, and Ø. Fischer, Phys. Rev. Lett. 75, 2754 共1995兲. 2 C. Renner, B. Revaz, K. Kadowaki, I. Maggio-Aprile, and Ø. Fischer, Phys. Rev. Lett. 80, 3606 共1998兲. 3 S. H. Pan, E. W. Hudson, A. K. Gupta, K.-W. Ng, H. Eisaki, S. Uchida, and J. C. Davis, Phys. Rev. Lett. 85, 1536 共2000兲. 4 S. H. Pan, J. P. O’Neal, R. L. Badzey, C. Chamon, H. Ding, J. R. Engelbrecht, Z. Wang, H. Eisaki, S. Uchida, A. K. Gupta, K.-W. Ng, E. W. Hudson, K. M. Lang, and J. C. Davis, Nature 共London兲 413, 282 共2001兲. 5 C. Howald, P. Fournier, and A. Kapitulnik, Phys. Rev. B 64, 100504共R兲 共2001兲. 6 K. M. Lang, V. Madhavan, J. E. Hoffman, E. W. Hudson, H. Eisaki, S. Uchida, and J. C. Davis, Nature 共London兲 415, 412 共2002兲. 7 K. K. Gomes, A. N. Pasupathy, A. Pushp, S. Ono, Y. Ando, and A. Yazdani, Nature 共London兲 447, 569 共2007兲. 8 A. N. Pasupathy, A. Pushp, K. K. Gomes, C. V. Parker, J. Wen, Z. Xu, G. Gu, S. Ono, Y. Ando, and A. Yazdani, Science 320, 196 共2008兲. 9 K. McElroy, D.-H. Lee, J. E. Hoffman, K. M. Lang, J. Lee, E. W. Hudson, H. Eisaki, S. Uchida, and J. C. Davis, Phys. Rev. Lett. 94, 197005 共2005兲. 10 J. W. Alldredge, J. Lee, K. McElroy, M. Wang, K. Fujita, Y. Kohsaka, C. Taylor, H. Eisaki, S. Uchida, P. J. Hirschfeld, and J. C. Davis, Nat. Phys. 4, 319 共2008兲.

12 O.

144506-15

PHYSICAL REVIEW B 80, 144506 共2009兲

KIMURA et al. 27

R. Ono, M. Cromar, R. Kautz, R. Soulen, J. Colwell, and W. Fogle, IEEE Trans. Magn. 23, 1670 共1987兲. 28 J. Martinis and R. Kautz, Phys. Rev. Lett. 63, 1507 共1989兲. 29 R. Kautz and J. Martinis, Phys. Rev. B 42, 9903 共1990兲. 30 G.-L. Ingold and Yu. V. Nazarov, in Single Charge Tunneling, edited by H. Grabert and M. Devoret 共Plenum, New York, 1992兲. 31 O. Naaman, Ph.D. thesis, University of California–San Diego, 2003. 32 T. A. Fulton and D. E. McCumber, Phys. Rev. 175, 585 共1968兲. 33 O. Naaman and R. C. Dynes, Solid State Commun. 129, 299 共2004兲. 34 P. W. Anderson, in Lectures on the Many-Body Problem, Ravello 1963, edited by E. R. Caianello 共Academic Press, New York, 1964兲, Vol. II, pp. 113–135. 35 Ø. Fischer, M. Kugler, I. Maggio-Aprile, C. Berthod, and C. Renner, Rev. Mod. Phys. 79, 353 共2007兲. 36 H. Kimura, R. P. Barber, Jr., S. Ono, Y. Ando, and R. C. Dynes, Phys. Rev. Lett. 101, 037002 共2008兲. 37 C. Renner, B. Revaz, J.-Y. Genoud, K. Kadowaki, and Ø. Fischer, Phys. Rev. Lett. 80, 149 共1998兲. 38 T. Cren, D. Roditchev, W. Sacks, J. Klein, J.-B. Moussy, C. Deville-Cavellin, and M. Laguës, Phys. Rev. Lett. 84, 147 共2000兲. 39 J. E. Hoffman, E. W. Hudson, K. M. Lang, V. Madhavan, H. Eisaki, S. Uchida, and J. C. Davis, Science 295, 466 共2002兲. 40 M. Vershinin, S. Misra, S. Ono, Y. Abe, Y. Ando, and A. Yazdani, Science 303, 1995 共2004兲. 41 J. F. Annett, N. Goldenfeld, and S. R. Renn, in Physical Properties of High Temperature Superconductors II, edited by D. M. Ginsberg 共World Scientific, Teaneck, NJ, 1990兲. 42 R. C. Dynes, F. Sharifi, and J. M. Valles, Jr., in Proceedings of the Conference on Lattice Effects in High-TC Superconductors, edited by Y. Bar-Yam, T. Egami, J. Mustre-Leon, and A. R. Bishop 共World Scientific, Singapore, 1992兲, pp. 299–308. 43 J. Lee, K. Fujita, K. McElroy, J. A. Slezak, M. Wang, Y. Aiura,

H. Bando, M. Ishikado, T. Masui, J.-X. Zhu, A. V. Balatsky, H. Eisaki, S. Uchida, and J. C. Davis, Nature 共London兲 442, 546 共2006兲. 44 F. C. Niestemski, S. Kunwar, S. Zhou, S. Li, H. Ding, Z. Wang, P. Dai, and V. Madhavan, Nature 共London兲 450, 1058 共2007兲. 45 S. Pilgram, T. M. Rice, and M. Sigrist, Phys. Rev. Lett. 97, 117003 共2006兲. 46 D. J. Scalapino, Nat. Phys. 2, 593 共2006兲. 47 A. Lanzara, P. V. Bogdanov, X. J. Zhou, S. A. Kellar, D. L. Feng, E. D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio, J.-I. Shimoyama, T. Noda, S. Uchida, Z. Hussain, and Z.-X. Shen, Nature 共London兲 412, 510 共2001兲. 48 V. J. Emery and S. A. Kivelson, Nature 共London兲 374, 434 共1995兲. 49 W. D. Wise, M. C. Boyer, K. Chatterjee, T. Kondo, T. Takeuchi, H. Ikuta, Y. Wang, and E. W. Hudson, Nat. Phys. 5, 213 共2009兲. 50 K. Tanaka, W. S. Lee, D. H. Lu, A. Fujimori, T. Fujii, Risdiana, I. Terasaki, D. J. Scalapino, T. P. Devereaux, Z. Hussain, and Z.-X. Shen, Science 314, 1910 共2006兲. 51 W. S. Lee, I. M. Vishik, K. Tanaka, D. H. Lu, T. Sasagawa, N. Nagaosa, T. P. Devereaux, Z. Hussain, and Z.-X. Shen, Nature 共London兲 450, 81 共2007兲. 52 M. C. Boyer, W. D. Wise, K. Chatterjee, M. Yi, T. Kondo, T. Takeuchi, H. Ikuta, and E. W. Hudson, Nat. Phys. 3, 802 共2007兲. 53 R. P. Vasquez, B. D. Hunt, and M. C. Foote, Appl. Phys. Lett. 53, 2692 共1988兲. 54 M. Gurvitch, J. M. Valles, Jr., A. M. Cucolo, R. C. Dynes, J. P. Garno, L. F. Schneemeyer, and J. V. Waszczak, Phys. Rev. Lett. 63, 1008 共1989兲. 55 J. M. Valles, Jr., R. C. Dynes, A. M. Cucolo, M. Gurvitch, L. F. Schneemeyer, J. P. Garno, and J. V. Waszczak, Phys. Rev. B 44, 11986 共1991兲. 56 A. G. Sun, A. Truscott, A. S. Katz, and R. C. Dynes, Phys. Rev. B 54, 6734 共1996兲. 57 A. Truscott, Ph.D. thesis, University of California–San Diego, 1999.

144506-16